University of Groningen
News devices Bounegru, Liliana
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date: 2019
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
Bounegru, L. (2019). News devices: how digital objects participate in news and research. University of Groningen.
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.
3. Narrating Networks: Network
Diagrams as Storytelling Devices in
Journalism
he empirical investigation of news devices undertaken in this
dissertation starts in the familiar terrain of journalistic stories.19 The
question around which this chapter is organised pertains to how we can account for narrative making or storytelling from a news device
perspective.
Digital news stories, these media products which we come across daily, materialise collective phenomena, and are experienced as multimodal constructions which combine written texts, images, audio, video and other resources according to journalistic genre conventions. In the context of these stories, studies have pointed to the role of digital objects such as graphics and visualisations as important techniques that news makers mobilise to support the narration of their stories (Segel & Heer, 2010), facilitate their
comprehension (Cairo, 2007; Schroeder, 2004), and attract reader attention (Utt & Pasternak, 2000). According to Latour (1990) images and graphics are powerful because they can at once stabilise phenomena and allow them to be transported across space, time and context, reasons for which he describes them as “immutable mobiles”.
In this chapter I contribute to understanding the role of visualisations in knowledge construction by exploring network graphs, hereafter called networks, in the context of journalistic storytelling. I focus on this type of digital object because over the past couple of decades, network graphs, understood as sets of nodes (or vertices) connected by edges (or links), have become an increasingly popular way to represent all kinds of collective
19 This chapter is based on an ongoing collaboration with Tommaso Venturini (French
National Center for Scientific Research), Jonathan Gray (King’s College London) and Mathieu Jacomy (Aalborg University) to examine network practices in the context of digital journalism, digital sociology and activism. An invitation to visit the Tow Center for Digital Journalism at Columbia University provided the initial impetus for this research. An article based on this study, of which I am the first author, was published in Digital Journalism, (Bounegru, Venturini, Gray, & Jacomy, 2017). Its development benefitted from input from participants in writing seminars led by Bruno Latour at the medialab at Sciences Po in Paris in December 2015 and March 2016. I am especially indebted to Paul Girard for his suggested clarifications the terminology used to describe the network story types. It also benefitted from feedback from participants in the research seminar of the University of Groningen’s Centre for Media and Journalism Studies in June 2015. This chapter provides an account of this research with a focus on the news device perspective.
phenomena. They have become a core aspect of visual culture in the digital age, illustrating and attesting to the complex webs of association around us. The metaphors and imagery of networks have exploded across many fields and their analytical capacities are said to have the potential to revolutionise
everything from medicine to markets to military intelligence. Advances in the computation and the visualisation of networks, the ubiquity of the internet as a network of computer networks and of the web as a network of hypertexts, have all contributed to this renaissance of networks (Barabási, 2002; Rieder, 2012; Watts, 2004).
Networks have also captured journalistic imagination. As journalists are increasingly organising their practices around the collection, analysis and narrativisation of structured and digital data (see, e.g., Anderson, 2018; Gray, Bounegru, & Chambers, 2012), the analytical and communicative potential of networks has received growing attention amongst journalism practitioners. Cases such as Josh On’s map of interlocking directorates of the most powerful US companies,20 Valdis Krebs’ mapping of the terrorist network around the
September 11 attacks (2002), Little Sis’ and Muckety’s maps of powerful people and organisations in the US,21 have opened up imagination about the
potential of network analysis and visualisation to make collective phenomena knowable journalistically.
While the mathematical properties of networks have received extensive attention for a long time (see, e.g., the research areas of graph theory and sociometry), in this chapter I follow the call of Espeland (2015) to scrutinise not just the analytical operations that make up quantification practices but also the circulation, interpretation, impact and narratives that numbers generate. Hence in this chapter I explore the stories about collective phenomena that network graph properties evoke in a series of journalistic projects and articles. Following the news device perspective described in Chapter 2, I take as an entry point to the study of journalistic storytelling not a journalism genre or media type but a defining object of digital culture, and examine its affordances
as a storytelling device.
I start by revisiting the link between journalism and storytelling, as well as how the narrative role of networks has been discussed so far. Next I introduce my corpus and research method and proceed to describe the five narrative
readings of networks I identified in a series of journalism examples. Finally, to support further research in this area I discuss methodological issues that I encountered and suggest directions for future study that I hope will advance and broaden research around this defining object of visual culture after the digital turn.
3.1 Journalism and Storytelling
Journalism has long been associated with practices of storytelling: the craft of rendering complex phenomena into narrative form (for an overview see, e.g., Kormelink & Costera Meijer, 2015; Zelizer, 2004). Schudson (1982) argues that the power of media stems not from the delivery of “facts”, but rather from the development of narratives and narrative forms, i.e. conventions by means of which sequences of actions or events are being reconstructed, ordered and presented. Bell (2005) writes that “journalists are professional storytellers of our age” (p. 397).
Journalism scholars have developed a rich tradition of studying the role of narratives in journalism. One strand of studies focuses on narrative styles, forms and conventions by means of which journalists decide what becomes news and how it is structured (see, e.g., Bell, 2007; Darnton, 1975; Tuchman, 1976). A famous such style is the “inverted pyramid” of narration which we often encounter in the news genre (Bell, 2007). Recognisable through its principle of organising and ordering facts from the most to the least important, the inverted pyramid organises narration around values such as recency,
novelty, human interest, conflict and focus on personalities (Bell, 2007; Dardenne, 2005).
capacity to create representations of the world that produce or reproduce ideological or value systems and are generative of character types such as the hero and the villain (see, e.g., Ettema & Glasser, 1988; Knight & Dean, 1982). For example, according to Ettema & Glasser (1998), in the US tradition of investigative journalism, narration often acts as an instrument for invoking morality in the service of judging civic vice, by portraying innocent citizens as victims of structural problems of the state.
In recent years the digitisation of journalism has been associated with renewed possibilities for narration, emphasising possibilities for more immersive, interactive, multimodal and participatory forms of storytelling (Pavlik, 2000, 2001). While some have cautioned about this optimistic interpretation of digital storytelling as journalism’s panacea (for a discussion see, e.g., Kormelink & Meijer, 2015), empirical investigation of storytelling with digital media is ongoing. For example, augmented reality (AR) as a storytelling medium has been understood to have the potential to increase audience engagement with news (Pavlik & Bridges, 2013). Twitter has been understood as a medium for participatory storytelling around news where citizens and journalists
collaboratively construct stories around news events (Papacharissi & Oliveira, 2012).
3.2 When Networks Meet Narratives
While the use of networks for data exploration and analysis has been studied extensively (see, e.g., Adamic & Adar, 2003; Andris et al., 2015), the narrative or storytelling potential of these digital objects is just beginning to receive more sustained attention from researchers.
Networks and narratives have recently come together in a number of different areas of research. One such area is that of information and communication technology and organisation studies, where narrative networks represent methodological devices for representing patterns and routines that emerge around the usage of information technologies (Pentland & Feldman, 2007;
Weeks, 2012). Another perhaps less expected direction of study is exemplified by the Narrative Networks programme set up by U.S. government’s Defense Advanced Research Projects Agency (DARPA). The programme aims to explore the relationships between narratives, human cognition and behaviour in the context of international security. Another more familiar area of inquiry that brings together these two concepts is narrative theory. In this context network analysis methods have been applied to the study of literary texts in attempts to develop quantitative approaches to enrich the study of narrative texts (see, e.g., Bearman & Stovel, 2000; Moretti, 2011; Sudhahar, Fazio, Franzosi, & Cristianini, 2015).
My interest differs from these other approaches in that I am predominantly interested in the narratives about collective phenomena that networks and their properties may elicit in situated storytelling practices. This is an area of
research that is just beginning to be explored (see, e.g., Bach et al., 2016; Suslik Spritzer et al., 2015; Venturini et al., 2017, 2018). Following the news device approach sensitivity towards the mutual shaping of technologies and practices (see Chapter 2), I am interested in how these digital objects are appropriated and used in situated storytelling practices, and, at the same time, in how their material affordances shape journalistic stories and translate aspects of the depicted collective phenomena. While the notion of affordance has different meanings and inflections, here I use it to account for both how the visual properties or features of networks invite particular narrative readings and to how these are activated in situated contexts of practice such as that of journalistic storytelling (for a discussion of the multiple ways in which this concept is used, see, e.g., Bucher & Helmond, 2018 and Nagy & Neff, 2015). To distinguish my approach from the various strands of work developed around the concept of narrative networks, I propose the complementary notions of network narratives and network stories. While distinctions between “narrative” and “story” have been drawn in narrative theory (for an
explanation see, e.g., Culler, 2001), in this chapter I use the terms “stories” and “narratives” interchangeably. The notion of network narrative or network story is intended to guide attention towards the narrative potential of these digital
objects, i.e. the ways in which narrative meaning may be constructed around network properties. By proposing this notion, I do not aim to suggest that network visualisations are narratives. Instead my aim is to draw attention, following Ryan (2004), to their potential to possess narrativity. Ryan characterises the distinction as follows:
The property of ‘being’ a narrative can be predicated on any semiotic object produced with the intent of evoking a narrative script in the mind of the audience. ‘Having narrativity,’ on the other hand, means being able to evoke such a script. In addition to life itself, pictures, music, or dance can have narrativity without being narratives in a literal sense. (p. 9)
3.3 How the Analysis Was Conducted
I started the analysis by building a collection of journalism pieces which use network diagrams and concepts.22 I collected 45 exemplary journalism pieces
from a number of different sources: interviews with journalists working in this area, online repositories of journalism stories such as those associated with the Data Journalism Awards competition23 and with the National Institute for
Computer-Assisted Reporting (NICAR),24 and specialty mailing lists such as
NICAR-L25 and Influence Mapping.26 These pieces belong to various
journalistic genres, from visual and investigative journalism, to special reports and interactives, to name just a few, and are published predominantly in digital formats. In these pieces, network graphics are integrated in multimedia
packages or used to illustrate pieces of writing. The stories and network graphics which I collected were not always accessed in their original medium of publication and hence some of the elements of context in which they were
22 For the purposes of this chapter I use the terms “network diagram”, “network visualisation” and “network graphic” interchangeably.
23 Accessible at: http://www.globaleditorsnetwork.org/programmes/data-journalism-awards/
24 Accessible at: https://www.ire.org/nicar/
originally embedded are absent. However, since these elements of context are essential to the interpretation of the narrative, I only considered in this analysis those network graphics for which sufficient elements of context were available. Given the illustrative nature of this analysis I ended up focusing on 13 of the 45 pieces that were identified in the collection that I created.
My device-centred approach to the construal of narrative meaning draws on a well-documented area of research, multimodal analysis (Jewitt, 2009b; Kress & van Leeuwen, 2001). Multimodality refers to “approaches that understand communication and representation to be more than about language, and which attend to the full range of communicational forms people use – image, gesture, gaze, posture and so on – and the relationships between them” (Jewitt, 2009a, p. 14). The network stories that I selected are realised through the interaction of multiple modes, from static and interactive diagrams, to photographs, pictograms, and written language, the latter being present in various forms, such as story headline, lead, body, graphic caption, labels, legend and
instructions. All these modes are meaningfully organised in the layout space of the journalistic piece. A mode is defined as “a socially shaped and culturally given resource for making meaning”, utilised in representation and
communication (Kress, 2009, p. 60).27
To facilitate the construal of narrative meaning from the point of view of the device, the network diagram, I used a model of analysis for the graphic
representation of networks, an area which is only beginning to be addressed in multimodal research (see, e.g., Bateman, Wildfeuer, & Hiippala, 2017). To construe narrative meaning from network graphics, I draw on Bertin’s (1983) semiotic model for the analysis of graphics, and on Venturini et al.’s (2015) framework for the visual analysis of networks. Bertin (1983) proposes that the reading of a graph largely consists of constructing correspondences around a central notion called the invariant, characterised by means of a series of visual variables, such as size, shape, colour and texture (p. 140-141). This reading occurs in stages, such as identifying the informational notions or concepts
27 For a more extensive discussion of semiotic modes see, e.g., Bateman (2011) and Hiippala
expressed through the graph and detecting the visual variables through which these notions are represented, before correspondences are being established between graph components. Crucial to the creation of such correspondences and thus to the construal of narrative meaning is a question, conscious or not, which guides the reader’s perception towards the relevant associations. Venturini et al. (2015) apply Bertin’s (1983) model to network graphics and propose a framework for constructing meaning around three visual variables of network graphs: node position, node size and node hue. According to this model, visual patterns corresponding to network properties such as clustering (Watts & Strogatz, 1998) and structural holes (Burt, 1992), may be read from the spatialisation of the network, i.e. the disposition and density of nodes in the graph. The property of an actor to be an authority, a hub or a bridge (Barabási, 2002; Granovetter, 1973), i.e. their centrality in the network, can be read from the position of nodes in the network, as well as from their size. Actor typology may be read from node colour or shape. In the next section, I illustrate how I applied these models to the analysis of multimodal journalistic stories with a network graphical component.
Another essential aspect in narrative construction and interpretation in our corpus is the genre of journalistic storytelling. Following the convention of this genre that the key points are presented at the beginning of the story,
particularly in the headline and the lead paragraph (Bell, 2007), where applicable, I have taken guidance in the formulation of the question that the network graphic invites the reader to ask, and hence of the story which it evokes, from the headline and lead of the journalistic piece, as well as from the graph elements addressed in textual form in the body of the story. Finally, the socio-cultural knowledge of the reader also plays an important role.
Drawing on these models, I take a number of steps in my analysis. Given my interest to explore narrative from the point of view of the digital object, the starting point of my analysis is to identify the notions, themes or concepts central to the journalism piece that are expressed through the network graphic, or what in multimodal research may be termed the diagrammatic mode
network visualisation which guide readers’ perceptions and by means of which narrative meaning is cued. The narrative view is then further specified and qualified by means of the textual elements in close proximity to the graph, most importantly its caption and the title, lead paragraph and body of the journalistic piece in which it is embedded, also mobilising more broadly our knowledge of the world and of the journalistic genre. I verbally formulate the narrative views which I construe and where applicable draw associations between them and the network concepts or properties which they deploy or with which they resonate.
To move from narrative readings to narrative reading types I use an emergent categorisation approach whereby categories of narrative readings are being construed by identifying repetitive story types or patterns in my collection of journalism pieces. Given the qualitative nature of my study focused on demonstrating the narrative potential of networks, I do not exhaustively analyse my collection of journalism pieces. Instead I focus on the construal of an illustrative set of narrative reading types, each of which is illustrated with the analysis of three examples.
3.4 Five Narrative Readings of Networks
This qualitative analysis shows that there are recurring narrative reading types that networks elicit and that these narrative readings resonate with or deploy a number of network properties or concepts. In this section I discuss five such examples of narrative readings and illustrate them with the analysis of three cases where they occur. As the analysis below shows, multiple narrative readings may be construed from a journalistic piece, which means that a journalism project or article may be discussed in relation to multiple categories. The construal of multiple narrative views occurs particularly in the case of pieces which deploy a mode which could be tentatively termed “dynamic diagram”. This composite mode incorporates a strong interactive component which helps to connect different narrative sequences and enables the narrative to progress. Unlike a static diagram, the dynamic network diagram enables the
user to explore and manipulate the graph display through a variety of interactive devices and techniques. Given the illustrative nature of my proposed typology of narrative readings, I do not undertake to exhaust all of the narrative readings of a journalistic project. Instead I focus on the ones that best illustrate the narrative views under examination.
3.4.1 Exploring Associations around Single Actors
In this category I grouped narrative views that depict the network of entities around a single actor. This category may be interpreted as evoking a particular type of network graph, often called “ego-network”. One typical characteristic of ego-networks is the depiction of relationships around a given social unit, referred to as the ego, resulting in “a mini-network or immediate
neighbourhood surrounding the ego” (Freeman, 1982, p. 291).
I also distinguish this category from the second narrative view which depicts key players, in that the nodes that play the role of the ego in this category of stories are not necessarily well-connected ones (authorities or hubs), which is where the emphasis lies in the “detecting key players” category in the next section. In borderline cases, where it has been difficult to place a story in one category or another, I returned to the written language mode and particularly to the journalism piece headline and lead as well as to the graphic headline to identify additional cues for the construal of the narrative view.
I will discuss three cases of narrative views which I have construed as “exploring associations around single actors” in this collection of journalism pieces. Across all these cases, crucial to the construal of the selected narrative reading is the interactive component of the dynamic diagrammatic mode. Interactive techniques such as clicking or mouse-over network elements to reveal details-on-demand, changing the appearance of the arrow cursor into a hand over active areas, implicit instruction through visualisation guides and explicit instruction through mouse-hover over elements of the graph, guide the reader to select single nodes in order to explore their networks of associations.
Thomson Reuters’ “Connected China” project (February 2013)28 is an interactive
multimedia website dedicated to “tracking thousands of people, institutions and connections that form China's elite power structure” by depicting
networks of familial, political and social ties amongst members of its governing structures. Of the multiple narrative readings elicited by the dynamic
diagrammatic mode, in this section I describe a prominent one, the exploration of associations around single actors. This narrative reading draws on multiple modes. In the written language and layout modes the importance of the ego is expressed through elements such as the guide to the reading of the
visualisation (see Figure 1), the textual explainers describing each individual in the network and the labels identifying them by name. In the diagrammatic mode, an important cue is the prominent positioning of the ego in the graph, its representation through a photograph of the individual whom it represents, and the disposition of the related nodes in a semicircle around it (see Figure 1(b)).
Figure 1: “Connected China”, Thomson Reuters, February 2013. (a)
Part five of the visualisation guide which provides instructions for navigating the “Social Power” section of the multimedia website. (b) Network visualisation depicting the network of associations around a
selected government official in China.
In the interactive micro-site “WESD Web of Connections” published by the
Statesman Journal (n.d.)29, the Willamette Educational Service District’s network
of connections is a multi-layered story about corruption, oversight and abuse undertaken by a regional education service agency in the United States.
Multiple modes participate in the construction of this narrative reading. In the default view of the interactive network graphic, the Willamette Educational Service District is set as the ego but users can move any other node to the centre of the graph by clicking on it. The written language mode, more specifically the graph title, instructions and labels, interacts with the dynamic diagrammatic mode and its spatial disposition of nodes and edges to guide the interpretation of the network around a single actor (Figure 2). The primary visual property through which the actor’s network is realised is the spatial
disposition of the ego at the centre of a radial graph, with edges radiating from it and connecting it to nodes representing suspect deals, administrative
structures and the individuals, organisations and addresses associated with these.
Figure 2: “WESD Web of Connections”, Statesman Journal, n.d. (a)
Special report title and instructions for navigating the interactive graphic. (b) Default graphic view depicting the Willamette Educational
Service District and its network of connections.
Finally, Washington Post’s “Top Secret America”,30 an investigative project
published on July 19, 2010, enables the user to explore the network of types of work conducted by and companies contracted by each of the 45 government
organisations that make up the national security program of the United States. This narrative reading is cued by the written language and dynamic
diagrammatic modes. An interactive device, an information box appearing upon hovering over elements of the visualisation, guides the reader to click on the name of an agency in order to explore its work areas and contractors (see Figure 3(a)). The ego is cued in the diagrammatic mode by its spatial
disposition at the centre of the graph, as well as the size of the node
representing it. Its network of entities is depicted through a series of circles of varying sizes and blocks of varying colours. The ego and its network of elements are further qualified in the written language mode through the graphic’s caption, the information box and the node label (see Figure 3(b)).
Figure 3: “Top Secret America”, Washington Post, July 19, 2010. (a)
Default view of the “Explore Connections” section of the investigation with information box inviting the reader to click on an entity to see its network of connections. (b) Network of contractors and areas of work
around selected government agency.
3.4.2 Detecting Key Players
In this category I grouped narrative views that depicted key actors based on the number of connections with other nodes. The focus of these network stories is on the density of connections around one or several central nodes. This category deploys the network property of power law distribution (Barabási, 2002). The Pareto or power law distribution of connectivity indicates the concentration of a large majority of connections around a small minority of nodes. Such nodes are called authorities or hubs, to show that they receive or spawn an unusually large number of links. Such a distribution of associations can be observed in the topology of the web, where a few webpages receive millions of incoming links, whereas close to 90% of all webpages receive ten or less incoming links (Barabási, 2002). This is also identified as a property of many other types of self-organizing networks in biology, economy and society.
Scientist’s “The Stem Cell Wars” special report31 on citation practices in stem
cell science, published on June 12, 2010. In this example the written language mode contributes to the narrative reading in multiple ways. The graphic caption guides the reading of the graphic around the detection of “influential players” (see Figure 4(a)). The information box “The strongest link” also cues this reading in its first sentence: “Shinya Yamanaka of Kyoto University in Japan is the dominant scientist in cellular reprogramming” (New Scientist, June 12, 2010; see Figure 4(b)). The most cited scientists, the Japanese Yamanaka and the U.S. based Jaenisch and Hochedlinger, are also discussed extensively in the body of the article. In the static network diagram, the node size as well as the depiction of edges as arrows pointing towards the most cited scientists support this narrative reading. The edges here represent citations received by papers authored by the scientists depicted as nodes in the graph.
Figure 4: “The Stem Cell Wars”, New Scientist. June 12, 2010. (a)
Article title and subtitle, graphic caption and network visualisation of citations in stem cell science. (b) Information box describing how the
network analysis was conducted, as well as its key findings about the influential players.
The article series entitled “Park Young-Joon at the Center of President Lee Myung-bak's Human Resources Network”, published on January 20, 2002, is a
JoongAng Ilbo investigation into the social network of senior officials around
South Korea’s president, Lee Myung-bak. In the written language mode, the headline of the article,32 the caption of the graphic,33 and the body of the
article all point towards the strength of associations of government official Park Young-joon to other top officials with links to the president. These links include whether they share the president’s hometown, whether they graduated from the same university, whether they served on the president’s campaign and a number of other business and political ties. In the static diagram the position of the node representing the most connected government official at the centre of the graph and the size of the icon representing him supports such a
narrative reading (Figure 5).
32 In translation: “In this social network among the 944 senior officials, Park Young-joon is
located at the center. An official is positioned toward the center when he or she has more links to President Lee”. Source: NICAR Stories Database, story no. 25691.
33 In translation: “In this social network among the 944 senior officials, Park Young-joon is
Figure 5: “Park Young-Joon at the Center of President Lee
Myung-Bak’s Human Resources Network”, JoongAng Ilbo, January 20, 2002. Article title, graphic caption and network visualisation depicting
government officials central to the president’s network.
My final example in this section is the “Flip Investigation”34 published by the
Sarasota Herald Tribune in 2009, examining real-estate fraud in the state of
Florida. While the dynamic diagrammatic mode invites multiple readings, in this section I focus on the “detecting key players” narrative view, which is prominently cued in the article series accompanying the interactive network visualisation. Here the key player is construed as a villain situated at the centre of a network of fraudulent transactions known as property flips. The narrative reading is signalled in the written language and photograph modes, with a full text article in the series, published on July 21, carrying the evocative title of “The King of the Sarasota Flip”, being dedicated to the investigation of the key player, Craig Adams (see Figure 6(a)).35 In the diagrammatic mode, the
narrative reading is realised through the default frame of the “Network” view of the interactive graph depicting Craig Adams at the centre of the flip deals network, comprising of other real-estate professionals, “flippers”, orchestrators
34 Accessible at: http://projects.heraldtribune.com/investigateflip/investigateflip.html 35 Accessible at: http://www.heraldtribune.com/article/20090721/ARTICLE/907211055
as well as victims (see Figure 6(b)). The topology of the network comes very close to the star-shaped “winner takes all” topology (Barabási, 2002), in which the central hub is connected to all the other nodes.
Figure 6: “The Flip Investigation”, Sarasota Herald Tribune, 2009. (a)
Title and picture from article in the Flip Investigation series investigating real estate agent Craig Adams. (b) Network visualisation depicting Craig Adams at the centre of a property-flipping network in
Florida.
3.4.3 Mapping Alliances and Oppositions
In this category I grouped narrative views that depict associations of nodes as well as the absence of associations between groupings of nodes. This category
deploys the network property of clustering, which measures the varying density of connections between nodes (Watts & Strogatz, 1998) as well as the property of structural holes (Burt, 1992). A cluster is thus a collection of nodes that are more densely connected among each other than to the rest of the network. A cluster is visually displayed by the spatial disposition of such nodes in
proximity to each other in a network graph. According to Burt (1992), the concept of structural hole refers to the “separation between nonredundant contacts” where nonredundant contacts are nodes that do not share
connections (p. 18). In the visual representation of a network, the clusters may be construed as alliances between actors and the absence of connections or structural holes may be construed as opposition or lack of allegiance (Venturini et al., 2015).
In what follows I will illustrate the construal of this narrative view with three examples from the political domain. Le Monde’s piece “2007-2011: La
Cartographie de la Blogosphère Politique”,36 is an interactive graphic
published on July 4, 2011 which represents the linkages amongst the French political blogosphere between 2007 and 2011. In the diagrammatic mode, the visual property through which the reading of alliances or coalitions between actors is realised is the spatial disposition of nodes in the graph into clusters. The clusters are identified through the density of associations amongst nodes, and through their colour (see Figure 7). Alliances may also be read not only between nodes but also between clusters. Such alliances can be read from the position of clusters in relation to one another. Clusters that are closer and share more links (such as the greens and the left bloggers’ clusters in the 2011 map) can be read as allies whereas the absence of links or the presence of structural holes between two clusters cues the reading of opposition (as in the case of the sparse connections between the extreme right on the one hand, and greens and left bloggers on the other in the 2011 map).
The journalistic genre convention that the most important information is presented at the beginning of the piece provides essential guidance, as the
36 In translation: “2007-2011: The Cartography of the Political Blogosphere”; accessible at: http://www.lemonde.fr/election-presidentielle-2012/visuel/2011/07/04/la-cartographie-de-la-blogosphere-politique_1544714_1471069.html
network of clusters is the default view of the graph. The alliances and
oppositions are further qualified through the written language mode. The title or headline of the piece identifies the graph as a map of the French political blogosphere. An interactive component, the navigation menu on the left-hand side of the graph identifies the political factions textually through their labels (“gauche”, “extreme gauche”, “ecologie politique”, “centre”, “droite”, and “extreme droite”)37 and pictorially through the colour of the icon associated
with each political faction, which in turn corresponds to the colour of the clusters on the map. The reading guide which accompanies the piece provides further guidance on the interpretation of the position of the nodes in relation to each other:
Ce type de placement permet de rendre visibles les dynamiques communautaires et le fait que certains sites échangent fortement entre eux et beaucoup moins avec les autres sites de la carte (qui seront donc plus éloignés) (Le Monde, 4 July 2011).38
37 In translation “left”, “extreme left”, “greens”, “right” and “extreme right” respectively. 38 In translation: “This type of disposition helps to make visible community dynamics and the fact that some sites exchange more links among themselves than with the other sites on the
Figure 7: “2007-2011: La Cartographie de la Blogosphère Politique”,
Le Monde. July 4, 2011. Title of journalistic piece and network graphic view of the French political blogosphere in 2011.
Global News’ piece “Visualising the Split on Toronto City Council”, published
on March 20, 2012,39 tells the story of the growing divergence between
Toronto council members between 2010 and 2012. As with the previous example, the alliances and oppositions are rendered in the diagrammatic mode through the spatial disposition of the nodes in separate clusters. The clusters are represented through different colours (red and blue) and are bridged by a few nodes in purple (see Figure 8). The narrative reading is further qualified as representing the alliances and oppositions around voting on the Toronto City Council through the written language mode, more specifically the headline of the article (“Visualizing the Split on Toronto City Council”), and the body of the article which identifies the factions as follows: “blue for the group around the mayor, red for the opposition, and purple for centrist or unaffiliated councillors” (Global News, 20 March 2012). The principle of association and
39 Archived version accessible at:
https://web.archive.org/web/20120507224159/http://www.globalnews.ca/topics/torontoco uncildiagrams/dec_2010/index.html
opposition is also identified in the written language mode: “Councillors who tend to vote together will be clustered closely together in the graphic” (Global
News, 20 March 2012). The coalitions and oppositions are calculated and
represented at different moments in time. This constitutes another narrative view type, which I discuss in the fourth category below.
Figure 8: “Visualizing the Split on Toronto City Council. Global News,
March 20, 2012. “December 2010” graphic panel depicting the group around the mayor (blue) and the opposition (red) and a few unaffiliated councillors positioned between the two clusters (purple).
Lastly, the Le Monde piece, “Mariage Gay: L'Opposition Soigne ses Amendements”, published on January 31, 2013,40 investigates coalitions
formed around amendments to the gay marriage law proposal in France. The diagrammatic mode signals the reading of coalitions through the spatial
40 Accessible at: http://www.lemonde.fr/societe/article/2013/01/31/mariage-pour-tous-l-opposition-soigne-ses-amendements_1825467_3224.html. In translation the title reads: “Gay
disposition of the nodes in several clusters in the interactive network graph which concludes the article. Political parties are represented through different colours and largely correspond to the clusters representing coalitions around the signing of legislative amendments (see Figure 9). The reading of coalitions is further specified in the headline and body of the article, as representing the alliances around the signing of amendments to the gay marriage law. The alliances are shown to follow ideological lines, with co-signing occurring largely within the boundaries of the political party, with the exception of the right-wing UMP (Union for a Popular Movement), which co-signs with independent parliament members as well as with the centrist UDI (Union of Democrats and Independents). The absence of connections between centre-left parties such as the RRDP (Radical, Republican, Democratic and Progressist) and ECOLO (Ecologists), as well as between centre-left and right-wing in the diagram can be construed as opposition. The principle of association between nodes in the graphic is also identified in the written language mode as follows:
Chaque cercle représente un parlementaire. Lorsqu'un député a signé ou cosigné un amendement avec un autre, un lien les relie. Les cercles sont ensuite placés de telle sorte que les cercles attachés les uns aux autres se rapprochent et que ceux qui ne sont pas attachés s'éloignent. La taille du cercle varie en fonction du nombre d'amendements signés par le député. (Le Monde, January 31, 2013)41
41 In translation: “Each circle represents a member of the Parliament. When a member has signed or co-signed an amendment with another, a link between them is established. The circles are then placed so that the circles linked to each other are closer. The size of the circle is proportional to the number of amendments signed by the member of Parliament”.
Figure 9: “Mariage Gay: L'Opposition Soigne ses Amendements”, Le
Monde. January 31, 2013. Instructions for navigating the graphic and network visualisation depicting coalitions around amendments to the
gay marriage law proposal in France.
3.4.4 Exploring the Evolution of Associations over Time In this category I grouped narrative views formed around a temporal
dimension and which show the transformation of associations of actors over time. This category deploys a property that is common to real-world networks, namely that they are dynamic systems whose composition and topology are subject to change over time (Barabási, 2002). In what follows I will illustrate this narrative view with three examples, two of which have been encountered in previous categories as well. In the case of all three examples, essential to the construal of this narrative reading are the composite modes of the dynamic diagram or the animated diagram (as is the case of the third example) with their interactive component. Interactivity is realised through devices such as a
navigation menu indicating different moments in time in the case of the first two examples and a time bar or progress bar in the case of the final example, all of which enable the reader to navigate across visualisation panels.
In Le Monde’s “2007-2011: La Cartographie de la Blogosphère Politique” (July 4, 2011), I identify a second narrative view, which I call “exploring the
evolution of associations over time”. The title or headline of the piece identifies the time period covered by the mapping: 2007 to 2011. An interactive device, the navigation menu (containing three buttons: “2011”, “2009”, and “2007”) and the caption of each visualisation panel (“le web politique militant en 2007”, 2009 and 2011 respectively) specify the particular year represented by the panel. The narrative view is also cued pictorially by the representation of network dynamics at two-year intervals through three maps, which the user can navigate from a menu at the top of the interactive graphic (see Figure 10).
Figure 10: “2007-2011: La Cartographie de la Blogosphère Politique”,
Le Monde. July 4, 2011. (a), (b) and (c). Network mapping of the French political blogosphere at three moments in time: 2007, 2009
The narrative view “exploring the evolution of associations over time” is similarly construed in the piece “Visualising the Split on Toronto City Council” (Global News, March 20, 2012). One important interactive element which cues this narrative reading is the navigation menu which enables the user to switch between representations of the network at various moments in time
(December 2010, May 2011, January 2012, February 2012, March-April 2012), as well as the caption of each graph, which indicates the moment in time which it depicts (see Figure 11). The longitudinal or temporal dimension is further qualified in the body of the article, where the key shifts in network dynamics at different moments in time are being analysed.
Figure 11: “Visualizing the Split on Toronto City Council. Global
News, March 20, 2012. (a), (b) and (c). Network visualisation of voting coalitions on the Toronto City Council at three moments in time:
December 2010, January 2012 and February 2012.
Twitter” (December 7, 2011),42 an investigation into the emergence and
correction of misinformation pertaining to the 2011 UK riots on Twitter, this narrative view is realised primarily with the participation of the dynamic diagrammatic mode, the written language mode and the layout mode. An interactive object, the time bar or progress bar present on the web page
dedicated to each rumour, enables the reader to follow the development of the rumour from its inception until its death. The diagrammatic mode signals the similarity of tweets through their spatial proximity in the graph, as well as their position in relation to the rumour (support, opposition, questioning and commentary), through the property of colour (see Figure 12). This narrative reading is further qualified through textual annotation highlighting key events in the unfolding of the rumour as well as through the intensity of the nodes’ colour, whereby lighter tones are used to represent more recent tweets and darker tones are used to represent older tweets.
42 Accessible at:
Figure 12: “How Riot Rumours Spread on Twitter”, Guardian,
December 7, 2011. (a), (b) and (c). Network diagram of clusters of tweets around the rumour “Rioters Attack a Children’s Hospital in Birmingham” at three different moments in the unfolding of the
3.4.5 Revealing Hidden Ties
In this category I grouped narrative views which depict hidden and potentially incriminating sequences of connections or paths between nodes. In this category the nodes represent (alleged) villains as well as their collaborators, whether individuals, businesses or government officials. The edges represent allegedly unethical, fraudulent and potentially incriminating relations. The painting by Marc Lombardi cited in the introduction of this chapter clearly falls in this category.
This category may be interpreted as loosely evoking the network property of “weak ties”. Whereas the nodes of a cluster are connected by strong ties, another essential concept in the study of social networks is that of weak ties (Granovetter, 1973). The notion is used to describe connections between nodes belonging to different clusters where nodes through which such connections are established act as bridges. Such nodes have been shown to play an essential role in various social activities through their ability to
transport information across the structural holes that separate clusters (Jensen et al., 2015). This narrative view should be distinguished from “exploring associations around single actors” and “detecting key players” in that, while selected actors are at the centre of such narrative views, the emphasis of the reading is not on one or several well-connected actors, but rather on the path of connections that ties the actors and the nature of these ties. I will illustrate this narrative view below with three examples.
The Kansas City Star’s “Terrorist Tentacles Know no Boundaries” (November
28, 2004) explores the ties between a global charity and multiple terrorist organisations and supporters, including terrorist Osama bin Laden. In the diagrammatic mode ties or connections are rendered through directed edges represented as arrows, all of which start from a single prominently sized node and point towards several other nodes, which can be construed as individuals or groups based on their icon depictions (see Figure 13). The nodes are further qualified through textual labels which identify them as the incriminated global charity IARA (Islamic African Relief Agency) and the terrorist organisations,
terrorism supporters and terrorists to which IARA is alleged to have ties, including Osama Bin Laden. The incriminating ties are further specified in the headline of the article ("Terrorist Tentacles Know no Boundaries") as well as in the body of the article where the ties are being listed:
At least eight connections between IARA and Osama bin Laden, his organisations or the Taliban; Two connections to Hamas, the
Palestinian terrorist organisations whose suicide bombings ravaged life in Israel; Connections to three other groups that long have been designated as terrorist organizations by federal authorities. (Kansas City
Star, November 28, 2004)
Figure 13: “Terrorist Tentacles Know no Boundaries”, Kansas City
Star, November 28, 2004. Tree-like network diagram depicting a global charity and its alleged terrorist ties.
(February 21, 2014),43 which investigates incriminating ties between a powerful
Southern Californian family and organisations in its area, both the diagrammatic and the written language modes participate in the narrative reading of incriminating ties. The incriminating ties or connections are rendered through arrows connecting two sets of nodes (see Figure 14). The nodes are qualified through textual labels as the members of the Calderon family and the organisations and businesses with which it has established incriminating relations. The nature of the relations represented by the edges is specified through colour and text: green arrows represent financial donations to campaigns, blue arrows represent legislative interventions or attempts thereof, and yellow arrows represent consultancy services. The headline of the piece (“The Calderon Family’s Connections”) further anchors this narrative reading. The subtitle of the article qualifies the ties as incriminating through the specification that the Calderon family is under investigation by the FBI. The nature of the ties is also cued by the body of the article, which provides further detail on the investigated connections between the family and various private and public actors in their region.
Figure 14: “The Calderon Family’s Connections”, Los Angeles Times,
February 21, 2014. Title, subtitle and interactive network graphic opening the journalistic piece.
Finally, I discuss the construal of incriminating ties in Organised Crime and Corruption Reporting Project (OCCRP)’s interactive piece “The Proxy Platform”, published on November 22, 2011.44 In this piece too, the path is
cued by the dynamic diagram and written languages modes. An interactive feature enables the reader to select a node and to visually highlight the
sequence of paths or connections to other nodes, as well as to reveal the name labels of these nodes, while the other elements of the graph are dimmed. The sequence of paths is composed of four different types of nodes. The
sequencing of the node types on a horizontal axis, inviting (but not restricting) the reading of connections between nodes from left to right further invites the construal of the notion of a path connecting different kinds of nodes (see Figure 15). The node types are specified through labels, from left to right, as: “proxies” (individuals running phantom companies on behalf of the real beneficiaries), “proxy companies” (companies set up to facilitate money
laundering), “banks” through which transactions flow between proxy companies and beneficiaries, and “beneficiary companies” which are on the receiving end of financial transactions). The interpretation of the sequences of connections as incriminating ties would be highly improbable from the
diagrammatic mode alone. It is the written language mode, more specifically the body of the article accompanying the interactive graph which describes the mechanisms of the money laundering system which spans Eastern Europe, Central and South America and Asia, that enables the identification of the sequences of paths between nodes as incriminating ties in a money laundering system.
Figure 15: “The Proxy Platform”, Organised Crime and Corruption
Reporting Project (OCCRP), November 22, 2011. Landing page of the journalistic piece depicting the interactive network graphic of actors
that make up the investigated money laundering system.
3.5 Conclusion
In this chapter I took a news device approach to the study of network
diagrams and their participation in journalism stories. Following this approach, the focus has been on how this digital object matters or makes a difference to narratives about collective phenomena in the context of journalism projects. I
described its contribution by highlighting five narrative reading patterns that networks elicit in my corpus of journalism pieces.
As colleagues and I discuss elsewhere, common metaphors that associate network graphs with “hairballs” or “spaghetti bowls” point to the absence of established practices for working with the narrative affordances of these objects, as well as to the limited literacy around reading them (Venturini, Bounegru, Jacomy, & Gray, 2017). Although illustrative rather than exhaustive, I hope that this vocabulary of narrative readings as well as the protocol for the construal of narrative meaning out of networks will contribute towards
elucidating the meaning-making capacities of these defining visual culture and knowledge-making devices in the digital age, and contribute towards
developing literacy to read them narratively.
Methodologically, the contribution of this chapter was to illustrate that existing analytical approaches such as multimodal analysis are well suited to support the analysis of narrative and communicative aspects of digital objects from a news device perspective. Drawing on this approach, I showed that forms of
storytelling that rely on network graphics are realised multimodally. It is in the interaction between modes such as static, dynamic or animated diagrams, written language, photographs, layout and pictograms that the specificity of this composite form of storytelling lies.
I show that the narrative readings of networks are co-produced through interactions between the material affordances of networks and the norms and values of the journalistic genres in which they are embedded. First, in keeping with Ryan’s (2005) proposition that good narratives think with their medium, that is that they take advantage of the affordances of the medium in which they are realised, I show that the material affordances of these digital objects shape narratives and how collective phenomena become widely knowable. Different narrative views tend to be cued by distinct visual attributes of network
diagrams such as node size and position, density of connections, node colour, and size and arrow-like depiction of edges. Narrative views deploy or evoke classic network concepts and properties, from ego-networks, to the power law
distribution, clustering, weak ties and the dynamic character of real-life networks. The production of narrative meaning revolves around relationality or associations and their properties, from topology, to density, to dynamic character or absence. For this reason, journalistic stories that I examined were often configured around the representation and exploration of structures, assemblages or collectives, be they political or power structures and fraud or corruption networks, to mention just a few.
In addition to their resonance with network concepts or properties, I also find these narrative readings to be shaped by some of the values and themes associated with the genres of news and investigative journalism. For example, the narrative readings constructed around the depiction of actors’ networks, key players and alliances and oppositions may be interpreted as resonating with news journalism values such as the focus on human interest, personalities, eliteness of news actors, and conflict (Bell, 2007; Dardenne, 2005). Narrative readings constructed around the exposure of incriminating ties resonate with conceptions of investigative journalism narratives to be acting as an instrument for invoking morality in the service of judging civic vice (Ettema & Glasser, 1998).
Returning to the multimodal realisation of these narrative views, besides being cued by particular visual attributes of networks, in all the examined cases the identification and qualification of the actors, connections between them, themes, temporal dimensions and other elements of the narrative reading would have been improbable without anchoring in the written language mode as well as context. The written language and layout modes present in our examples in the form of headlines, article bodies, graphic labels, captions and guides, are so important in the construal of narrative meaning that in this analysis I resolved to eliminate multiple examples of network graphics that had been used in the context of journalism pieces in cases where their original context of publication was unavailable to me and I was thus unable to unambiguously construe narrative meaning. The journalism genre and
particularly its convention that the most important information is contained in the headline and lead of an article, and the subjectivity and socio-cultural
knowledge of the reader more broadly, also played an important role in guiding the interpretation of narrative meaning. For this reason, I place emphasis in this analysis on the description of the process by means of which I have
construed each narrative reading.
Given the small collection of examples I make no claims of
comprehensiveness or representativeness of the narrative views which I identified. While for the purposes of this chapter I limited myself to the discussion of five narrative views that exploit or resonate with network concepts or properties that have captured the popular network imagination today, I believe that other narrative readings may also be construed (for example around the concept of “spheres of influence”) and invite researchers to conduct further work in this area. To facilitate such work I make available together with this chapter the full collection of examples of journalism pieces deploying network visualisations or concepts and encourage other researchers to expand it.45
I also do not claim the boundaries between the identified narrative reading types to be clear-cut. In fact, I encounter in this analysis several borderline cases where, on the basis of the diagrammatic mode alone, the narrative meaning can be equally construed along several narrative views. In these cases, I anchor or qualify the narrative reading by taking into account information present in other elements of the journalistic piece, particularly the headline, lead and body of the article.
I would also like to note that the narrative views which I identified are not mutually exclusive. In fact, in this collection I was often able to construe multiple narrative views from a single journalism piece. These function as building blocks for the broader journalistic narrative. This was particularly the case of pieces drawing on the dynamic diagrammatic mode, where these views were sequences of a larger narrative through which the reader can progress thanks to a strong interactive component realised through a diversity of
45 The list is accessible here:
techniques and devices. In addition to this, given the aim to qualitatively explore the narrative potential of networks, I have not exhaustively analysed all the narrative views elicited by networks that make up the larger narrative developed in a journalism piece. Instead I selected and limited my focus to illustrating a few narrative views that resonate with classic network concepts and properties. I do however consider the interplay between different narrative sequences in the context of the broader narrative of a single journalism piece to be important and invite researchers to study it.
Finally, to further situate the findings of this chapter, I will address some final broader issues pertaining to the analytical perspective used. I will suggest that while studies that focus on journalistic representations such as the one described in this chapter are necessary and insightful, the frame of study should be extended beyond journalistic representations in their original context of publication to also explore their performativity across different sites and contexts.
In this chapter I focused on the interpretation of these graphics in their narrow context of publication and in doing so I left out both the practices of
production of such images, and their circulation and reception dynamics. Such production practices need to be accounted for because visual displays reflect the conventions of the professional practices that create them and the affordances and constraints of the visualisation techniques used, as much as they reflect the characteristics of the depicted phenomena themselves (Bastide, 1990; Lynch, 1985). As far as circulation dynamics are concerned, journalistic images and visualisations may take on different roles, functions and
interpretations depending on who they reach and the various situations in which they are received and used.
In the context of the social study of art, Becker (1982) proposes the concept of “art worlds” to emphasise the need to expand the analytical frame beyond the familiar art works themselves to how they are circulated and constructed through “the network of people whose cooperative activity, organised via their joint knowledge of conventional means of doing things, produces the kind of
art works that art world is noted for” (p. 10). In the next chapter I shift my focus from news works to the journalism coding work which increasingly underpins the production of journalistic projects, news tools and
infrastructures. In doing so I also shift focus from digital objects as sites of narrative and communicative acts, to digital objects as sites of platformisation of news work (Chapter 4) and commodification of news audiences (Chapter 5).
